Linebender released the Fearless SIMD 1.0 library for Rust

The team of developers at Linebender has released the production version of the Fearless SIMD 1.0 library, designed for safe work with vector processor instructions in the Rust language. The tool completely eliminates the use of unsafe blocks when performing SIMD operations, replacing them with safe type conversions and specific compiler functions.

The main innovation of version 1.0 is the #[simd] macro, which automates the process of multiversioning code. This allows developers to compile the same logic under different sets of processor instructions, such as SSE, AVX2, or NEON, without writing duplicate code. The library provides performance comparable to low-level implementations while maintaining strict memory safety guarantees characteristic of the Rust ecosystem.

The release includes two types of API: precise and fast. The precise variant guarantees exact compliance with the behavior of hardware instructions, while fast allows the compiler to apply additional optimizations that may slightly change the calculation result in favor of speed. All components of the library have been tested for compatibility with the current versions of the rustc compiler.

Migrating to Fearless SIMD reduces the number of segmentation errors and logical vulnerabilities that occur when manually managing vector registers. The tool is focused on high-load tasks where maximum data throughput is required while maintaining strict typing and code security.